From owner-freebsd-hackers Sat Dec 30 17:33:49 1995 Return-Path: owner-hackers Received: (from root@localhost) by freefall.freebsd.org (8.7.3/8.7.3) id RAA28663 for hackers-outgoing; Sat, 30 Dec 1995 17:33:49 -0800 (PST) Received: from localhost.cybercomm.net (sl-039.sl.cybercomm.net [199.171.196.167]) by freefall.freebsd.org (8.7.3/8.7.3) with SMTP id RAA28658 for ; Sat, 30 Dec 1995 17:33:42 -0800 (PST) Received: from sl-039.sl.cybercomm.net (localhost [127.0.0.1]) by localhost.cybercomm.net (8.6.12/8.6.12) with SMTP id UAA01326; Sat, 30 Dec 1995 20:33:11 -0500 Date: Sat, 30 Dec 1995 20:33:09 -0500 (EST) From: Sujal Patel X-Sender: smpatel@sl-039.sl.cybercomm.net To: Joerg Wunsch cc: FreeBSD Hackers Subject: Re: /dev/io In-Reply-To: <199512302111.WAA01418@uriah.heep.sax.de> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-hackers@freebsd.org Precedence: bulk On Sat, 30 Dec 1995, J Wunsch wrote: > Is there any good reason why access to /dev/io is allowed for members > of group kmem? Note that this will give any setgid kmem program > _full_ (read and write) access to the io registers (since open() is > the only important action for this device). While we're on the topic of /dev/io, I have a couple of comments about it. It seems unintuitive to offer a process IO permission by opening a device (especially since opening it with RDONLY or RDWR is irrelevant). I know that NetBSD changed this implementation after 1.0 to support i386_iopl(), i386_get_ioperm(), and i386_set_ioperm() calls. Also, Linux uses a set of calls very similar to NetBSD's. To me it seems like it would be more "standard" if FreeBSD supported an implementation like NetBSD's. It would also pave the way for the implementation of IO permission bitmaps (if they are ever needed). Changing it would probably be a little more secure, and I can't even think of any programs that would break if /dev/io was removed? Any comments? I'm interested in hearing about this... Sujal